Knowledge catalogue
arXiv:2609.15998v1 Announce Type: new Abstract: Every large language model (LLM) has behavioral traits and moral preferences that comprise its character. Whether by design or as an emergent property o
arXiv:2609.14796v1 Announce Type: new Abstract: The threat that AI persuasion poses to human control has been acknowledged in the literature, but not yet systematically studied. Now that persuasion at
opencl: choose the MoE expert matmul by batch size for speculative decoding/MTP (#27637) opencl: gate the prebuilt q4_0 MoE GEMM on routing count opencl: stop writing zeros into the padded MoE activat
arXiv:2609.15849v1 Announce Type: cross Abstract: Can LLMs reason through new information like humans, or do they merely retrieve cached opinions? This is critical for silicon sampling, where LLM pers
arXiv:2609.15706v1 Announce Type: new Abstract: Vision-language models (VLMs) are increasingly used to extract structured fields from business documents, yet most evaluations report accuracy on clean
Good Start Labs experiments show that AI training with gameplay can lead to transferable skills: a version trained on a railroad simulation outperformed peers in financial research. Inspired by Twitch
arXiv:2609.15383v1 Announce Type: cross Abstract: Language model safety is typically evaluated one interaction at a time. We show that a weaker, unaligned model can split a harmful task into benign-lo
arXiv:2609.14839v1 Announce Type: cross Abstract: The integration of Large Language Models (LLMs) into automated code optimization introduces a critical reliability risk we term the Efficiency Halluci
arXiv:2609.15422v1 Announce Type: new Abstract: AI agents are provisioned the same as employee-owned hosts in many enterprise settings with a static credential set fixed at deployment which includes a
arXiv:2609.15134v1 Announce Type: new Abstract: Computer-use agents increasingly interact with browsers, terminals, file systems, and external services, introducing safety risks that emerge through ru
arXiv:2609.13637v1 Announce Type: new Abstract: Persistent agents need evaluations that distinguish identity facts they can recall from those they express and enact. We introduce PAI-Bench, a provider
arXiv:2609.13936v1 Announce Type: new Abstract: Datasets that ship automatically generated feature annotations invite a question rarely asked of them: would a human agree with those labels? This repor
arXiv:2609.15207v1 Announce Type: new Abstract: Generative AI writing assistants and the Large Language Models (LLMs) that power them are increasingly part of how voters gather information before elec
arXiv:2609.15038v1 Announce Type: new Abstract: Language-model agents are increasingly used as synthetic people in surveys and social simulations, yet their apparent realism is often judged from popul
THE SANDBOX WAS A PROP! OpenAI Turned Off the Guardrails, Left a Door to the Internet, and Then Sold the Hugging Face Breach as “Rogue AI” It is time to understand how you were lied to and by whom. In
arXiv:2609.15494v1 Announce Type: new Abstract: Recent investigations of the July 2026 OpenAI--Hugging Face incident motivate two questions about agent behavior under task failure: when an assigned ta
arXiv:2609.13348v1 Announce Type: cross Abstract: This paper presents ViFA-Council, a three-stage multi-agent framework that employs multiple large language models (LLMs) to tackle two culturally comp
I know I know, it's great, we know. I've been working on tweaking inference engines for a week now and it's been one shotting most of my vague prompts without any issues. It will even write tests and
For all the focus on hard math, we should also have a bank of historical mysteries and cyphers that we can throw models at. I know a few historians (like https://generativehistory.substack.com/) that
A new paper (I'm not a coauthor BTW -- I just found it interesting) argues, basically, that RSI1 is not on the horizon2, because current (at the time the study was done) agents cannot do open-ended ML
arXiv:2609.12331v1 Announce Type: new Abstract: Simulated students generated by computational models provide a practical way to evaluate tutoring strategies and pedagogical approaches used by human an
DeepSeek V4.1 Flash is now available on Together AI’s model hub (https://www.together.ai/models/deepseek-v4-1-flash). The release includes the DeepSeek V4.1 Flash API with updated pricing, performance
Hey everyone, Giving AI agents access to MCP tools or shell commands is awesome, but prompt injection is still a huge headache. If an agent reads an untrusted document containing `ignore previous inst
A second way in which Dario’s letter does not hold up to close inspection. (See my newsletter for the first.) In his Pacing AI letter, Dario claims that RSI is starting to happen. The problem with thi
In September 2026, Simon Willison noted that while Anthropic had previously taken down PyPI packages, OpenAI's recent operation against RubyGems was notably more forceful. The tweet referenced an inve
I tried using deepseek v4.1 flash max subagents with astra extra high as an orchestrator. It was working perfectly with 1 until it suddenly stopped responding or doing work after 3 hours. Then trying
Fast meets open. 🚀 Qwen3.8-27B is now running on @cerebras with rapid inference. Try it now! Qwen3.8-27B is now live at Cerebras speed. The dense, open-weight model from @Alibaba_Qwen scores 34 on the
arXiv:2609.11180v1 Announce Type: new Abstract: Large language model (LLM) coding agents constantly decide whether a version satisfies a constraint such as ^1.2.3 or >=2.0,1.2 means >=1.3.0) traps eve
agreed, one of the biggest bottlenecks right now is my own ability to internalize the information that these models convey the latest models (astra, fable 5.1) are either way too verbose, or try too h
arXiv:2609.10611v1 Announce Type: cross Abstract: Membership inference attacks (MIAs) have emerged as critical tools for auditing privacy risks in large language models (LLMs), aiming to determine whe
Learn how to build and deploy an MCP App with interactive HTML widgets on Amazon Bedrock AgentCore. Because MCP Apps is a host-agnostic standard, the same server delivers the same rich experience acro
Datasette 1.0a39 and 0.65.4 security releases Today we're releasing two new security patch versions of Datasette: 1.0a39 and 0.65.4 - one for the current alpha series and one for the stable 0.65.x fam
arXiv:2603.17094v2 Announce Type: replace Abstract: Simulating human conversations using large language models (LLMs) has emerged as a scalable methodology for modeling human social interaction. This
arXiv:2606.21844v2 Announce Type: replace Abstract: As AI systems integrate into online spaces, differentiating them from humans in conversations is increasingly important. We present Inverse Turing B
arXiv:2604.00280v2 Announce Type: replace-cross Abstract: Formal specifications play a central role in ensuring software reliability, yet automatically synthesizing high-quality specifications remains
arXiv:2609.09604v1 Announce Type: cross Abstract: On August 2, 2026, the obligations of Article 50 of the EU AI Act took effect, requiring generative AI providers to mark the content their systems pro
arXiv:2606.16723v2 Announce Type: replace Abstract: Large language model (LLM) agents increasingly take actions (screening applicants, recommending credit, triaging patients), yet fairness for LLMs is
arXiv:2609.05818v1 Announce Type: new Abstract: We introduce ABLE, a benchmark for evaluating LLM agents' ability to use biological AI models (BAIMs), such as ProteinMPNN and AlphaFold3, in dual-use p
arXiv:2609.08248v1 Announce Type: new Abstract: Modern industrial ads ranking stacks are increasingly bottlenecked not by model capacity or training compute, but by the throughput of human ML iteratio
Another brilliant paper from Meta. This one is worth reading if you work on production-grade ranking or recommendation systems. Meta deployed an autonomous agent that runs the ML iteration cycle acros
Individual hackers are now sustaining campaigns that would have needed many skilled operators a year ago, because artificial intelligence has absorbed the labor that used to set state-backed teams apa
arXiv:2609.08861v1 Announce Type: new Abstract: Benchmark scores are a central currency in model releases: they inform purchasing decisions, shape public trust, and influence policy. Yet, a key assump
arXiv:2609.05802v1 Announce Type: cross Abstract: Large language models answering questions over multi-page documents are expected to cite the supporting pages, yet supplied citations are sometimes in
arXiv:2609.09410v1 Announce Type: new Abstract: While autonomous agents have made significant strides in 'deep research' by iteratively navigating the open web to synthesize information, real-world pr
arXiv:2609.09468v1 Announce Type: new Abstract: Can an agent learn a numerical search strategy through executable practice and then transfer that strategy as text? We study low-budget black-box optimi
arXiv:2601.13591v3 Announce Type: replace Abstract: Recent LLM-based data agents aim to automate data science tasks ranging from data analysis to deep learning. However, the open-ended nature of real-
arXiv:2609.09790v1 Announce Type: cross Abstract: Large Multimodal Models (LMMs) large-scale deployment in industrial warehouse settings specifically necessitates that models exhibit human-expert-leve
arXiv:2609.09984v1 Announce Type: new Abstract: Understanding the historical allocation and distribution of research funding advances our knowledge of how scientific research is supported across field
arXiv:2609.09070v1 Announce Type: cross Abstract: Clinical AI evaluation should encompass diagnosis and management after adaptive information gathering. We compared Doctorina, eight physicians and fou
arXiv:2601.09721v2 Announce Type: replace-cross Abstract: Consumer artificial intelligence chatbots are now accessed by hundreds of millions of users seeking health information, yet systematic evaluat
arXiv:2609.06226v1 Announce Type: cross Abstract: Large language models (LLMs) have recently made substantial progress in formal proof generation, yet presenting distinctive challenges in cryptographi
arXiv:2609.08236v1 Announce Type: new Abstract: Automatic safety judges -- systems such as Llama Guard or a GPT-4o grading prompt that decide whether a model's reply is harmful -- produce the numbers
arXiv:2609.06229v1 Announce Type: cross Abstract: Vulnerability discovery is becoming an important ability of large language model (LLM) agents: agents that silently miss real defects leave critical s
The new DeepSeek V4.1 Flash model is mindblowing - back on top of the open-source model leaderboard and extremely cheap. It has a lot of very smart ways to be efficient and highly capable so I made a
arXiv:2609.10315v1 Announce Type: cross Abstract: Reinforcement learning with verifiable rewards (RLVR) has advanced language-model reasoning in domains such as mathematics and code, where objective a
arXiv:2609.05435v1 Announce Type: new Abstract: Modern language agents are expected to operate over long horizons: they ask follow-up questions, reuse worked examples, handle tool feedback, and adapt
Also note that well aligned models are not sufficient to solve prompt injection by themselves yet. That said, we have solved prompt injection in practice when layering our latest models with prompt in
Astra is awesome, but if you're using it specifically to parse and extract a massive volume of documents, you will break the bank (10c+ per page on parsing and extraction!) We've added a native LlamaP
I have been a dwarfstar fan for awhile and I really liked glm 5.3 flash but needed it to be materially faster to feel good using it. In the screenshot you can see the outcome of using the model with a
arXiv:2609.06058v1 Announce Type: new Abstract: As vision-language models (VLMs) become increasingly capable and are deployed in consequential real-world settings, they must evaluate evidence independ